Disruptive Risk Identification Technology in Textile Intelligent Manufacturing Using Heterogeneous Graph Neural Network

The increasing deployment of wireless sensing, industrial Internet of Things (IIoT), and electromagnetic information transmission technologies in textile intelligent manufacturing has generated massive multi-source heterogeneous data, making accurate disruptive risk identification essential for ensuring reliable system operation and efficient information propagation. However, conventional approaches struggle to capture latent correlations and cross-domain propagation paths, particularly under small-sample conditions. To address this issue, this paper proposes a disruptive risk identification method based on a heterogeneous graph neural network (HGNN). A heterogeneous knowledge graph is constructed to model entities and relationships among equipment, processes, materials, and quality information, while a dynamic update mechanism enables real-time graph evolution. Meta-path-guided multi-head graph attention is employed to achieve adaptive semantic aggregation and cross-type information propagation, and graph-level representation learning combined with anomaly scoring accurately identifies disruptive sources and transmission paths. Experimental results demonstrate that the proposed method achieves an average recognition accuracy of 86.6% across five representative meta-paths. In small-sample scenarios with only 2.3% abnormal samples, the recall and F1-score reach 81.3% and 83.0%, respectively, significantly outperforming existing heterogeneous graph models. The proposed framework provides an effective solution for intelligent risk warning in textile manufacturing and offers valuable methodological support for reliable information processing and decision-making in electromagnetic-enabled industrial sensing and communication environments.